Has the concept of using a hand checked simple theorem prover to validate more complex theorem provers been explored before?

Question

More specifically, has anyone used a chain of theorem provers to validate a highly evolved theorem prover, starting with a very simple hand checked prover such that each new theorem prover is used to validate a somewhat more complex theorem prover, rinse repeat?

This would be something along the lines of induction, I guess. In that you'd have just as much confidence in the complex prover at the end as the hand checked prover you started with.

If it has been explored (wouldn't be surprised), a pointer towards more resources would be welcome.